• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1; RUN: llc -march=amdgcn -mcpu=bonaire -show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=CI -check-prefix=ALL %s
2; RUN: llc -march=amdgcn -mcpu=carrizo --show-mc-encoding < %s | FileCheck -check-prefix=GCN -check-prefix=VI -check-prefix=ALL %s
3; RUN: llc -march=amdgcn -mcpu=bonaire -mtriple=amdgcn-unknown-amdhsa < %s -mattr=-flat-for-global | FileCheck -check-prefix=GCNHSA -check-prefix=CIHSA -check-prefix=ALL %s
4; RUN: llc -march=amdgcn -mcpu=carrizo -mtriple=amdgcn-unknown-amdhsa -mattr=-flat-for-global < %s | FileCheck -check-prefix=GCNHSA -check-prefix=VIHSA -check-prefix=ALL %s
5
6; FIXME: align on alloca seems to be ignored for private_segment_alignment
7
8; ALL-LABEL: {{^}}large_alloca_compute_shader:
9
10; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
11; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD0
12; GCN-DAG: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
13; GCN-DAG: ; fixup A - offset: 4, value: SCRATCH_RSRC_DWORD1
14; GCN-DAG: s_mov_b32 s{{[0-9]+}}, -1
15; CI-DAG: s_mov_b32 s{{[0-9]+}}, 0xe8f000
16; VI-DAG: s_mov_b32 s{{[0-9]+}}, 0xe80000
17
18
19; GCNHSA: .amd_kernel_code_t
20
21; GCNHSA: compute_pgm_rsrc2_scratch_en = 1
22; GCNHSA: compute_pgm_rsrc2_user_sgpr = 8
23; GCNHSA: compute_pgm_rsrc2_tgid_x_en = 1
24; GCNHSA: compute_pgm_rsrc2_tgid_y_en = 0
25; GCNHSA: compute_pgm_rsrc2_tgid_z_en = 0
26; GCNHSA: compute_pgm_rsrc2_tg_size_en = 0
27; GCNHSA: compute_pgm_rsrc2_tidig_comp_cnt = 0
28
29; GCNHSA: enable_sgpr_private_segment_buffer = 1
30; GCNHSA: enable_sgpr_dispatch_ptr = 0
31; GCNHSA: enable_sgpr_queue_ptr = 0
32; GCNHSA: enable_sgpr_kernarg_segment_ptr = 1
33; GCNHSA: enable_sgpr_dispatch_id = 0
34; GCNHSA: enable_sgpr_flat_scratch_init = 1
35; GCNHSA: enable_sgpr_private_segment_size = 0
36; GCNHSA: enable_sgpr_grid_workgroup_count_x = 0
37; GCNHSA: enable_sgpr_grid_workgroup_count_y = 0
38; GCNHSA: enable_sgpr_grid_workgroup_count_z = 0
39; GCNHSA: workitem_private_segment_byte_size = 32772
40; GCNHSA: private_segment_alignment = 4
41; GCNHSA: .end_amd_kernel_code_t
42
43
44; GCNHSA: buffer_store_dword {{v[0-9]+}}, {{v[0-9]+}}, s[0:3], s9 offen
45; GCNHSA: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, s[0:3], s9 offen
46
47; Scratch size = alloca size + emergency stack slot
48; ALL: ; ScratchSize: 32772
49define void @large_alloca_compute_shader(i32 %x, i32 %y) #0 {
50  %large = alloca [8192 x i32], align 4
51  %gep = getelementptr [8192 x i32], [8192 x i32]* %large, i32 0, i32 8191
52  store volatile i32 %x, i32* %gep
53  %gep1 = getelementptr [8192 x i32], [8192 x i32]* %large, i32 0, i32 %y
54  %val = load volatile i32, i32* %gep1
55  store volatile i32 %val, i32 addrspace(1)* undef
56  ret void
57}
58
59attributes #0 = { nounwind  }
60